Dell Fluid File System v3

8/18/2019 Dell Fluid File System v3

1/17

Dell Fluid File System v3.0 A Dell Technology White Paper

Dell Product Group

June 2013


2/17

DELL TECHNOLOGY WHITE PAPER: Dell Fluid File System v3.0

THIS TECHNOLOGY WHITE PAPER IS FOR INFORMATIONAL PURPOSES ONLY, AND MAY CONTAINTYPOGRAPHICAL ERRORS AND TECHNICAL INACCURACIES. THE CONTENT IS PROVIDED AS IS,

WITHOUT EXPRESS OR IMPLIED WARRANTIES OF ANY KIND.

© 2013 Dell Inc. All rights reserved. Reproduction of this material in any manner whatsoever without

the express written permission of Dell Inc. is strictly forbidden. For more information, contact Dell.

Dell, the DELL logo, and the DELL badge, EqualLogic, Compellent, PowerVault and NetVault are

trademarks of Dell Inc. Microsoft and Windows are either trademarks or registered trademarks of

Microsoft Corporation in the United States and/or other countries. Symantec, BackupExec, and

NetBackup are trademarks of Symantec Corporation or its affiliates in the U.S. and other countries.

CommVault and Simpana are registered trademarks of CommVault Systems, Inc. Other trademarksand trade names may be used in this document to refer to either the entities claiming the marks and

names or their products. Dell Inc. disclaims any proprietary interest in trademarks and trade names

other than its own.

June 2013, version 3.0


3/17


Contents

1 Abstract ............................................................................................................................................................... 4

2 Dell Fluid File System Architecture ........... .......... ........... .......... ........... .......... .......... ........... .......... ........... ........ 5

2.1 Layered Approach to FluidFS Logical Architecture ........................................................................... 5

2.1.1

Network Load-Balancing Layer ....................................................................................................... 5

2.1.2 File Protocol Access ........................................................................................................................... 6

2.1.3 NAS Volumes ....................................................................................................................................... 6

2.1.4

NAS Pool .............................................................................................................................................. 6

2.1.5 Block Device LUNs .............................................................................................................................. 7

2.2 The Physical Building Blocks ........... .......... ........... .......... ........... .......... .......... ........... .......... ........... ......... 7

2.2.1 NAS Appliance ...................................................................................................................................... 7

2.2.2 Back-end SAN Storage ...................................................................................................................... 8

2.3

Summary ................................................................................................................................................... 8

3 Data Integrity ....................................................................................................................................................... 10

3.1 Cache Mirroring and Metadata ........................................................................................................... 10

3.2 High availability ...................................................................................................................................... 10

3.3 Summary ................................................................................................................................................. 10

4

Data Protection ................................................................................................................................................. 11

4.1 Snapshots ................................................................................................................................................. 11

4.2 Backups .................................................................................................................................................... 11

4.3

Replication ...............................................................................................................................................12

4.4 Summary ..................................................................................................................................................12

5 Data Reduction .................................................................................................................................................13

5.1 Data Deduplication .................................................................................................................................13

5.2 Compression ...........................................................................................................................................13

5.3 Summary ................................................................................................................................................. 14

6 Fluid File System Solutions ......... ........... ........... .......... ........... .......... .......... ........... .......... ........... .......... ........... 15

6.1

Dell Compellent ...................................................................................................................................... 15

6.1.1

FS8600 ................................................................................................................................................ 15

6.2

Dell EqualLogic ...................................................................................................................................... 16

6.2.1 FS76x0 ................................................................................................................................................ 16

7 Summary ............................................................................................................................................................ 17

7.1 Additional Reading .................................................................................................................................. 17


4/17

4


1 Abstract

Traditional approaches to handling file data growth have proven to be costly, hard to manage, and

difficult to scale effectively and efficiently. Dell™ Fluid File System (FluidFS) is designed to go beyondthe limitations of traditional file systems with a flexible architecture that enables organizations to scale

out and scale up non-disruptively. Hence, it addresses organizational challenges by allowing them to

gain control of their data, reduce complexity, and meet growing data demands over time.

The FluidFS architecture is an open-standards based network attached storage (NAS) file system that

supports industry standard protocols including NFS v4 and CIFS/SMB v2.1. It provides innovative

features proving high availability, performance, efficient data management, data integrity, and data

protection. As a core component of the Dell Fluid Data architecture, FluidFS brings differentiated

value to the various Dell storage offerings including the Dell Compellent FS8600 and FS8610 as well

as the Dell EqualLogic FS7600 and FS7610. Continue reading to learn about additional features and

enhancements that make it unique from other NAS products available in the market today.


5/17

5


2 Dell Fluid File System Architecture

The relentless growth of unstructured file data is accelerating the need for network file storage

systems. Organizations coping with data growth are confronted with several challenges:

Data silos prevent easy access to vital business information.

Data migration, backup, and disaster recovery are complex, consuming administrative time

and resources.

Meeting data growth by deploying more and more storage systems increases both the

administrative burden and capital expenditure at a time when businesses need to run lean.

Traditional file systems have scalability limitations that make them unwieldy for organizations

with rapidly expanding file data.

FluidFS is an enterprise-class, fully distributed file system that provides customers with the tools

necessary to manage file data in an efficient and simple manner. The underlying software

architecture leverages a symmetric clustering model with distributed metadata, native load balancing,

advanced caching capabilities and a rich set of enterprise-class features. FluidFS removes the

scalability limitations such as limited volume size associated with traditional file systems and supports

high capacity, performance-intensive workloads via scaling up (adding capacity to the system) and by

scaling out (adding nodes, or performance, to the system).

FluidFS operates across a symmetric cluster of purpose-built NAS controllers (housed in pairs within a

2U appliance), which interface over a fabric to shared back-end SAN storage, the EqualLogic PS

Series and Compellent Storage Center in particular.

The following sections provide descriptions of the logical and physical layers of FluidFS.

2.1 Layered Approach to FluidFS Logical

Architecture

The FluidFS architecture approaches layering with

performance in mind. The layered architecture presents a

traditional file system to network clients while performing a

range of special functions at the back end. The specific goal

of this design is to utilize all available resources at the

network, server and disk levels to support the fastest

possible response times.

The logical layers that comprise a FluidFS system and

influence system performance are described below.

2.1.1 Network Load-Balancing Layer

FluidFS presents multiple network ports from multiple

controllers to the client network. To simplify access and

administration, FluidFS supports Virtual IP addresses and physical distribution of client sessions is

Network Load-Balancing

CIFS/SMB Share or NFS Exports

NAS Volumes

NAS Pool

Block Device LUNs


6/17

6


managed by a built-in load balancing capability. The load is balanced between the interfaces within a

controller using one of two industry standard algorithms — Adaptive Load Balancing (ALB) or Link

Aggregation Protocol (LACP). ALB is a MAC address-based balancing mechanism that does not

require any configuration on the network switch. LACP, which is also known as 802.3ad, is another

available option and is supported by most major manufacturers ’ managed switches; some

configuration is required on the switch.

For client access within the same layer 2 network, FluidFS balances client connections across all

available network interfaces within the cluster based on client MAC addresses. When load balancing

across a routed network boundary, FluidFS utilizes DNS Round Robin in addition to the MAC-based

address lookup. Each NAS controller can access and serve all data stored in the FluidFS system.

Figure 1 shows load balancing in a Fluid File System cluster on a single layer 2 network.

Figure 1 - Load Balancing in a FluidFS Cluster

2.1.2 File Protocol Access

FluidFS supports the CIFS/SMB and NFS protocols. All shares and exports configured by the system

administrator are available from all NAS controllers at all times in a manner that is completely

transparent to the client.

2.1.3 NAS Volumes

NAS volumes are virtual entities that provide policy-based management of snapshots, replication,

quotas, deduplication, backup and security style as appropriate for different workloads. NAS volumes

can be created on the fly and can shrink or grow, non-disruptively, in capacity up to the physical

limits of the back-end storage.

2.1.4 NAS Pool

At its core, FluidFS creates a single distributed file system which spans across all SAN block device

LUNs that are presented to the system. This layer, represented as the NAS Pool, provides file system

services to the NAS volumes defined by the administrator in an optimized manner. All NAS controllers

actively participate in the FluidFS distributed file system and can write or read any data housed within

the file system.


7/17

7


2.1.5 Block Device LUNs

FluidFS integrates with best-in-class Dell SAN solutions EqualLogic and Compellent. LUN resources

are automatically provisioned to FluidFS by defining an arbitrary amount of desired NAS capacity.

Figure 2 depicts

Figure 2 – Logical Constructs of a FluidFS Cluster

2.2 The Physical Building Blocks

FluidFS interacts with two main hardware elements:

NAS gateway appliance(s)

Back-end SAN storage fabric and array

The minimum hardware configuration of a FluidFS system is a single NAS appliance and a single SAN

“unit” consisting of RAID controllers and disks.

This configuration can scale capacity by adding additional disks to the SAN and scale performance by

adding additional NAS appliances or SAN storage controllers. The NAS system can scale capacity and

performance independently and online, without disrupting system availability.

2.2.1 NAS Appliance

Two active-active, hot-swappable NAS controllers are housed side by side within a 2U FluidFS NAS

appliance. A 40Gb/s mid-plane between the two controllers enables cache mirroring and write-back

operations. Internal hard drives and backup batteries protect metadata and ensure data integrity.

The appliance is connected to a block-based storage system – a Compellent or EqualLogic SAN – via

8Gb Fibre Channel or 10Gb iSCSI. The cluster can scale out to multiple NAS appliances according to

client workload characteristics and business needs. To achieve data distribution and maintain high

availability, each controller pair in a cluster has access to all other controller pairs in the cluster


8/17

8


through a dedicated and redundant interconnect network. Each FluidFS NAS controller can also read

from and write to any file in any file system. Figure 3 represents the logical architecture of a FluidFS

cluster with 2 NAS appliances.

Figure 3 – A Dual NAS Appliance Cluster

The number of NAS appliances clustered together in a FluidFS system dictates the aggregate

compute, cache and client network bandwidth available on the system. However, proper system

sizing for performance must also consider the performance of the SAN back-end.

2.2.2 Back-end SAN Storage

FluidFS is integrated with Dell’s leading block-level SAN storage systems, either Compellent or

EqualLogic. Whether purchased as a complete system (including the SAN block device) or as an add-on to existing Compellent or EqualLogic components, FluidFS can be configured to meet

organizational and needs.

One hundred percent of the back-end storage system can be assigned to provide block-level LUNs

for the NAS controllers. Or the SAN resources can be shared with other applications that utilize direct

block access.

Key configuration options on the back-end storage system include:

Number of SAN controllers

Number and type of disks

RAID level and number of tiers

The specific configuration will dictate the overall performance of the SAN. For more information on

FluidFS performance, read the Dell FluidFS Architecture for System Performance White Paper .

2.3 Summary

Every file accessed by a client on the network passes through the logical layers described above.

Starting at the Network layer, the client is load-balanced to a specific NAS controller and continues

http://www.dellstorage.com/WorkArea/DownloadAsset.aspx?id=3504http://www.dellstorage.com/WorkArea/DownloadAsset.aspx?id=3504http://www.dellstorage.com/WorkArea/DownloadAsset.aspx?id=3504http://www.dellstorage.com/WorkArea/DownloadAsset.aspx?id=3504


9/17

9


through the protocol layer via a CIFS/SMB share or NFS export to the NAS volume layer, the

distributed file system layer and culminating with a write or read operation by the block device.

The configuration and number of NAS controllers combined with back-end storage system

components defines the system characteristics in terms of capacity and performance. The FluidFS

scale-up and scale-out capabilities allow the system’s initial configuration to evolve over time to

support the changing needs of the environment.


10/17

10


3 Data Integrity

FluidFS provides a series of mechanisms to provide a high level of integrity and system resiliency fordata at rest and in transit. This section details those mechanisms.

3.1 Cache Mirroring and Metadata

During normal operation, the write cache, which includes the data and the metadata, is mirrored

between the controller pairs in the FluidFS cluster. Additionally, important metadata is journaled to

back-end storage capacity by a journaling process that runs continuously. This ensures file system

consistency, in the event of a failure.

When a controller fails, cache mirroring is not possible, so the surviving controller immediately

journals its data and metadata to the storage presented by the back-end storage subsystem. This

ensures that the file system remains consistent and that all data is protected in case of additionalfailures.

In the event of a power outage, the write cache is journaled to a temporary staging location within the

controller. This ensures that all I/O in flight remains consistent and prevents data loss. In addition,

this ensures data consistency is not compromised, irrespective of the duration of the outage, as seen

by traditional battery backed systems, which will lose data if power is lost before the write completes.

These mechanisms provide resiliency and redundancy for a multitude of failure scenarios to ensure

that data is always intact and accessible.

3.2 High availability

In a FluidFS cluster, any single controller can fail without affecting data availability or causing data loss

— even if write operations were in flight. Cross-cluster reliability is achieved through a variety of

mechanisms, including a high speed cluster interconnect, write cache mirroring, failsafe journaling,

and data integrity checks to ensure data store consistency.

FluidFS monitors the health of the NAS appliance, including temperature and power conditions, to

ensure cluster reliability and maximize data availability in cases of hardware or software failures. If

failures occur, hardware components in the storage subsystem are redundant and hot- swappable.

Each controller receives its power from the power grid and a dedicated backup power supply (BPS),

which is regularly monitored to ensure that the BPS maintains a minimum level of power for normal

operation. The BPS has sufficient battery power to allow the controllers to execute their shutdown

procedures and use the cache as NVRAM. The BPS also provides enough time to write all the data

from the cache to disk in the event of a loss of power.

3.3 Summary

FluidFS data integrity features provide both redundancy and high availability for several different

scenarios. By working together you can rest assure that your data will always be available.


11/17

11


4 Data Protection

FluidFS enables data protection within a single system, across systems, and to external NAS

repositories. This section will discuss some of the features and benefits of using FluidFS to store andprotect data long-term.

4.1 Snapshots

Snapshots provide the first level of data protection by providing the ability to recover data instantly

and are an integral part of FluidFS. Each NAS volume has its own snapshot policy to allow flexibility

and to optimize space management. The administrator has the ability to schedule, point-in-time,

automatic snapshots as frequently as every five minutes and can initiate manual snapshots at any

time.

FluidFS incorporates redirect-on-write snapshots, instead of the copy-on-write solutions typical of

other file systems. Redirect-on-write requires only one I/O operation, thereby preventing

performance degradation. Figure 4 shows the redirect-on-write mechanism and the different stages

of the snapshot process.

Before Snapshot After Snapshot After Data Modification

ActiveProductionFile System



Snapshot Snapshot

--

--

--

--

--

-

-

-

--

-

-

-

--

-

--

-

-

-

-

--

-

--

--

--

--

--

-

-

-

--

-

-

-

--

-

--

-

-

-

-

--

-

A B C D A B C D A B C D D’

Blocks of Data Blocks of Data Blocks of Data

Figure 4 - Redirect-on-Write Snapshot Process

Snapshots are available to users as a read-only copy of the file system, which allows them to restore

their documents in a simple manner without administrator intervention. Administrators also can

easily restore very large data sets (terabyte scale) as a whole to a particular point in time. This

eliminates long file copies and the need for free space for the recovery process.

4.2 Backups

The Dell Fluid File System supports standard backup software using Network Data Management

Protocol (NDMP) version 4 with no changes required to existing backup workflows. Dell has

partnered with industry leaders to provide comprehensive backup solutions that integrate with Fluid

File System. Currently supported backup software includes:


12/17

12


Dell Quest NetVault Backup 9

CommVault® Simpana® 9.x

Symantec™ BackupExec™ 7.x

Symantec™ NetBackup™ 2010R3 and 2012

Tivoli Storage Manager 6.3

Please refer to the product specification sheets for the latest information on backup software

supported with FluidFS products.

4.3 Replication

FluidFS allows fast and reliable snapshot-based replication of any number of volumes to a partner.

After the initial synchronization, only incremental changes are replicated, which improves network

bandwidth utilization. This replication is native to FluidFS and does not require any additional

hardware. The data is always consistent on the partner site and available as read-only.

In addition to data, NAS configurations (volumes, exports, etc.) are replicated. This reduces

administrative burden and enables continuous access to data in the event of a disaster or site failureto assure business continuity.

Replication is bi-directional, meaning that the same system can host both source and destination

volumes. In addition, the direction can be reversed without requiring a full resynchronization. FluidFS

also supports “one-to-many” and “many-to-one” replication between NAS systems using unique

volumes. Figure 5 shows the supported replication options.

Figure 5 - FluidFS Replication

4.4 Summary

The use of FluidFS data protection features enables users to prevent data loss and to ensure

continued operations for most storage environments, with limited impact on users and applications.


13/17

13


5 Data Reduction

FluidFS v3 provides policy based data reduction capability which can be used to reduce data footprint and thus maximize return on investment in Dell Storage. Data reduction polices are set on a per

volume basis. FluidFS data reduction is designed to align with typical life cycle management policies

of unstructured data. Data reduction is available as part of FluidFS v3 and does not require separate

hardware or licensing. This section will discuss some of the features and benefits of using data

reduction and how it can benefit your data.

5.1 Data Deduplication

FluidFS deduplication utilizes a post-write process. Files are written to the FluidFS appliance in their

native form and then deduplicated once they meet a pre-specified set of reduction criteria. The

administrator can set deduplication policies for each NAS volume based on the last time a file was

accessed or modified, and can also schedule deduplication activity for a certain time window. For

instance, a data set that will be rarely or never read or modified again, such as backup, can be set to

be immediately reduced, and to do so after hours. Hot data that will typically experience many

writes, such as home shares, can be scheduled to deduplication after a longer period of inactivity.

When a file is modified, FluidFS will only rehydrate the portion of the file with write activity,

minimizing I/O.

Fluid Data Reduction uses a variable-block sliding-window strategy. It crawls each file, looking for

matching chunks of data sized between 64KB and 192KB with a general goal of 128KB. The reduction

engine then eliminates redundancies, creates an object block map and stores that information back

on the NAS volume.

Under normal circumstances, FluidFS will not optimize data that has not been accessed or modified

in fewer than 30 days. However, there will be times where data will rarely be accessed after it is

written to the file system. This is the case with backup/archive data or large image stores as one

might find in a PACS system. For these instances, a setting called “archive mode” allows the NAS

administrator to set a policy that enable data deduplication immediately, without waiting for the

minimum active data policy to run.

5.2 Compression

Compression is the second sub feature of the core data reduction capabilities of FluidFS. Compression of

files in FluidFS is accomplished using the LZPS algorithm. This was developed by Dell and uses less

CPU overhead than traditional compression algorithms. This cuts down on the overhead associated withreading/writing small files.

Any NAS volume can be configured for only deduplication or deduplication plus compression. The

NAS administrator can schedule when to run data reduction and on which NAS volumes. As more

NAS appliances or controller pairs are added to a NAS cluster and as the new controllers own more

data, not only does the NAS cluster scale in file performance, but also scales in the speed at which

data can be reduced.


14/17

14


Once the data reduction job starts, all files meeting the data reduction criteria are processed for data

reduction. Since only the files that are not actively used are reduced, FluidFS does not incur any

performance impact for active data. For read access to data that has been reduced, the blocks

requested by the client are rehydrated or changed to native form on the fly by FluidFS and passed to

the client. Due to rehydration, read performance can be impacted. Reads do not alter how the

blocks are stored on the disk- reduced blocks remain in reduced state. Any blocks within reduced

data that are modified are stored in native form. When such files with modified blocks meet the data

reduction policies, the modified blocks will be again reduced and stored in reduced form.

5.3 Summary

As unstructured data continues to grow, data reduction will become increasingly important. By

taking advantage of FluidFS data reduction features it will be possible to store large amounts of data

efficiently all while following typical data lifecycle management policies.


15/17

15


6 Fluid File System Solutions

FluidFS is being implemented as a core component in a number of Dell storage solutions that serve

the needs of different customer scenarios and data workloads. Each solution has unique feature,function and value propositions, and each is differentiated by the type of controllers being used in a

FluidFS cluster and the architecture of the underlying block-based back-end storage subsystem. This

approach to design, coupled with the distributed and clustered architecture results in solutions that

can easily leverage future technology enhancements.

Current solution that incorporate FluidFS are:

Dell Compellent FS8600 – Unified Fibre Channel and iSCSI storage platforms that offer high-

performance, scale out capability and efficient use of storage resources through features

such as thin provisioning and Automated Tiered Storage technology that moves data to the

optimal storage tier and/or RAID level.

Dell EqualLogic FS76x0 – Unified iSCSI storage platforms that offer high-performance scale

out capability. The scale out architecture native to both FluidFS and EqualLogic storage

systems provides linear performance scalability in line with capacity growth.

6.1 Dell Compellent

Dell Compellent provides an extremely agile platform for constantly evolving block and file storage

requirements. Performance and capacity scale non-disruptively to accommodate growing storage

needs without forcing a platform rip-and-replace.

6.1.1 FS8600

The Compellent FS8600 scale-out architecture

supports a single namespace across as many as

four appliances and capacity that expands up to

2PB of manageable space with two Storage Center

arrays. The inherent resilience of the FS8600

provides robust data protection without adding

complexity.

For additional information refer to the product

datasheet located on DellStorage.com.

http://www.dellstorage.com/compellent/nas/http://www.dellstorage.com/compellent/nas/http://www.dellstorage.com/compellent/nas/http://www.dellstorage.com/compellent/nas/


16/17

16


6.2 Dell EqualLogic

Dell EqualLogic NAS platforms are high-performance

solutions that enable organizations to easily configure

and manage iSCSI, SMB, and NFS storage from a single

interface. As storage needs change, block and file

capacity can be modified without disrupting

applications and storage systems. A single file system

can be expanded up to the capacity of the EqualLogic back end. NAS service can be configured and added

to EqualLogic arrays that have been deployed quickly and efficiently. EqualLogic NAS products include a

file-based snapshot capability (separate from iSCSI snapshots). Users can restore previous versions of files

from a directory of these snapshots themselves, without contacting IT.

6.2.1 FS76x0

The FS7600 and FS7610 are the second-generation NAS appliances in the EqualLogic product line.

They are based on FluidFS v2 and work with all EqualLogic PS arrays to provide highly available scale-

out NAS and unified storage solutions. These solutions deliver the core benefits of the EqualLogic

platform, including peer scaling, ease-of-use and all inclusive software licensing and they enable both

performance and capacity to scale without SAN or application downtime.

The FS7600 has 1GbE connectivity to the SAN and client network and the FS7610 has 10GbE

connectivity to the SAN and client network. Both appliances work with existing PS Series arrays to help

provide comprehensive unified storage and NAS functionality for midsize and smaller deployments.

With the PS storage back end, block and file storage capacity can expand up to 509TB, and

performance can be increased by scaling out across two FS76x0 appliances. Most notably, with

FluidFS, a NAS Volume or share can scale to the capacity of the backend storage allocated to file

storage, all within a single namespace.

For additional information refer to the product datasheet located on DellStorage.com.

http://www.dellstorage.com/equallogic/nas/http://www.dellstorage.com/equallogic/nas/http://www.dellstorage.com/equallogic/nas/http://www.dellstorage.com/equallogic/nas/


17/17

17


7 Summary

Dell’s Fluid File System adds non-disruptive, scale out and scale up NAS capabilities to Dell’s primary

storage products: Compellent and EqualLogic. Additionally, it removes the scalability limitationsassociated with traditional monolithic NAS architectures. It also provides innovative features proving

high availability, performance, efficient data management, data integrity, and data protection. With

the added data reduction features of FluidFS v3.0 handling file data growth is now within the grasp of

organizations of all sizes.

7.1 Additional Reading

For more detailed information on FluidFS performance, read the Dell FluidFS Architecture for System

Performance White Paper.

For more detailed information on each of the Dell Storage Products mentioned in this document,

please review the product datasheets located on DellStorage.com.
http://www.dellstorage.com/WorkArea/DownloadAsset.aspx?id=3504http://www.dellstorage.com/WorkArea/DownloadAsset.aspx?id=3504http://www.dellstorage.com/WorkArea/DownloadAsset.aspx?id=3504http://www.dellstorage.com/WorkArea/DownloadAsset.aspx?id=3504http://www.dellstorage.com/http://www.dellstorage.com/http://www.dellstorage.com/http://www.dellstorage.com/http://www.dellstorage.com/WorkArea/DownloadAsset.aspx?id=3504http://www.dellstorage.com/WorkArea/DownloadAsset.aspx?id=3504

Dell Fluid File System v3

Documents

Transcript of Dell Fluid File System v3